Research On Localization For Mobile Robot Based On LADAR

Mobile robot self-localization is to judge the robot's relative position from the environment and its attitude according to varieties of information, which consists of absolute localization and relative localization and is critical in robot navigation and control. This paper mainly studies on Mobile robot self-localization.At first, the related characteristics of mobile robot platform based is analyzed, on the basic of it a new autonomous AGV platform on laser radar is developed independently. The coordinate axes model,motion model and sensor model of mobile robot are established then. The common methods of mobile robot global positioning is analyzed and compared.A global positioning method for mobile robot based on Markov is advanced to realize mobile robot absolute localization. This method can be divided into two stages: time updating and observation updating. In the time updating phase, a time update method based on frequency domain processing is proposed to achieve non integer multiple raster translation of reliable image and to get the preliminary estimates of the reliability of mobile robot's pose. In the observation updating phase,in order to achieve the likelihood calculation without data association, a method based on Gaussian kernel smooth is put forward and the reliability of mobile robot's pose is estimated. At last, several simulation examples on the platform of Matlab is designed to verify the validity of the global positioning method based on Markov.A pose tracking method for mobile robot is proposed to achieve the relative localization and this is a point to point correspondence type laser scanning matching method. Firstly, the robot's environment information at continuous and different times is obtained and the mathematical model of related positioning parameters is established, using the previous state as a reference scan and the current state as the current scan. Secondly, the incremental rotation and translation are obtained by running 1D Fourier-Mellin three times continuously. At last, several simulation examples on the platform of Matlab and many experiments of pose tracking in real situation are conducted. The advantages of this method on positioning accuracy, computational complexity and the adaptability of complex environment are highlighted by compared with that of other ways.